Lecture motor YS share - Neuroinformaticskiper/Lecture_motor_YS.pdf · 2018-11-19 ·...

Systems Neuroscience Nov. 13, 2018

Yulia SandamirskayaINI, UZH/ETHZyuliasa@ethz.ch

Motor Systems

Why is motor system important?

• All behavior is directly or indirectly related to activity in the motor system

➡story of the mollusk ➡most of the neural system, in the end, serves movement➡movement is prerequisite for ANY perception and learning➡“thinking” and cognition use the same structures as physical movement

• Without the neuronal structures that evolved to actively generate behavior, we could not experience sensation, think, reason, problem solve, read, write, and do mental math, and we would not be able to communicate our thoughts and abilities to anyone.

What’s the brain good for?

A story of Ascidiella Scabra ….

Overview of motor systems

• Spinal reflexes

• Motor, pre-motor, and all the other cortical regions

• Cortical-subcortical-thalamo-cortical systems-Involving basal ganglia-Involving pons and cerebellum

• Corticospinal and corticobulbar tracts

Skeletal Muscles (vs. smooth muscles)

- striated (striped) appearance because they are comprised of muscle fibers

- move through a pull action (contraction)

- work in pairs with a reciprocal muscle (bicep contracts & triceps relaxes)

- stimulated by a Motor Neuron

The Biceps and Triceps

Anatomy of the muscle ➡ striated muscles are made of muscle fibers that have

two parts, outer and inner:

Outer fiber = extrafusal fiber

Inner fiber = intrafusal fiberThis represents only one muscle

fiber - a muscle

has many fibers

Wrapped around the intrafusal fiber is a sensory nerve that picks up the sensation of stretch.

Neurophysiology of the muscle Outer fiber = extrafusal fiber

Inner fiber = intrafusal fiber

Gamma Motor Neuron

Alpha motor neuron

• gamma motor neuron that synapses on the intrafusal fiber • alpha motor neuron synapses on the extrafusal fibers

- the neural link between the alpha motor neuron and the muscle fiber is called the neuromuscular junction.

- one alpha motor neuron can stimulate numerous fibers ( “Motor unit”)

Muscle-Spindle Feedback Circuit

Some facts

•The ratio between the alpha motor neuron and the number of muscles fibers it innervates is associated with the degree of dexterity needed in the movement- high ratio (1:150) = contraction of large muscles - low ratio (1: 10) = contraction of small muscles needed for fine

movements

Motor Homunculus is related to the number of alpha motor neurons needed to innervate muscles of various regions of our body.

Comparing the Anatomy of the CNS with the Anatomy of the Neuromuscular Junction

Motor Unit

•Alpha Motor Neuron•Muscle Fiber•Endplate•NT is Acetylcholine•Nicotinic Receptors•Calcium enters•Endplate Potential (EPP)•Muscle Contraction or Muscle Action Potential & movement

CNS Synapse

Presynaptic NeuronPostsynaptic NeuronDendriteMany different NTsMany different receptorsSodium entersEPSPAction Potential & release of NT

How is limb position maintained?

•Involuntary movement (i.e. posture): ➡continual contraction and relaxation of the muscles in our feet and calves.

•Voluntary movement:➡ Stretch of the intrafusal fiber causes contraction of

the extrafusal fiber via alpha motor neuron. ➡ Keeping the movement at this position requires a

direct signal from the brain.

The distinction between voluntary and involuntary might be illusive (posture control and visual feedback, loss of proprioception, “automatic” goal-directed movements…).

Automatic Maintenance of Limb Position (cont.)

“Force control”

Automatic Maintenance of Limb Position (cont.)

“Force control”

Intrafusal motor neurons

Remember, muscles work in pairs, so if one contracts the other relaxes

This is referred to as reciprocal innervation.

What if both muscles contracted at the same time?

Reciprocal innervation of Antagonistic Muscles

The elicidation of a Stretch Reflex

Motor Neurons•Alpha Motor Neuron is the Final Common Path for all movement. Movement can be generated from:

- sensory signals in the muscle spindle like the stretch reflex

- sensory signals from skin as in the pain withdrawal response

- involuntary signals from the brainstem for posture, keeping us upright without conscious attention

- signals from the brain for voluntary movementgoal directed, flexible, adaptive, learned

But, regardless of where the signal originates, all movement is the result of activity in the alpha motor

neuron – making this the Final Common Path

The two divisions of the dorsolateral motor pathway

Corticospinal tract•Origins:

- primary motor cortex (MI)- premotor cortex- supplemental motor cortex- anterior paracentral gyrus- parietal lobe (including SI) - cingulate gyrus

• Collaterals: small percentage of corticospinal neurons

- midbrain (primarily red nucleus)- trigeminal nuclei- pontine nucle

Corticospinal tract

Termination in spinal cord:

- mostly laminae 3-7, few in ventral horn and laminae 1-2; - mostly innervating interneurons;- some innervation of alpha motor neurons.

Neurotransmitter:

- glutamate and/or aspartate

Piramidal tract origin

Corticobulbar tracts A. control over facial muscles; bilateral input to motor neurons controlling muscles in upper face, but contralateral input to motor neurons controlling lower face (in humans, not sure about rodents)

B. control over muscles of mastication: motor trigeminal, and RF

C. control over external eye muscles: input comes from frontal and parietal eye fields, rather than from MI; projection to midbrain and paramedian pontine RF

D. control over tongue: hypoglossal and RF

E. control over swallowing reflexes: nucleus ambiguus and RF

Voluntary movement: Instructions from the cortex

•Dorsolateral Prefrontal Cortex: directs movement of our limbs (as in reaching) and movements of our fingers.

•Actual signal for movement must go through pre-motor cortex, then motor cortex.

•From motor cortex, signal travels down spinal cord eventually reaching the alpha motor neuron.

•BUT, the instructions for this movement ultimately comes from our Parietal lobe, which receives sensory input.

Cortical input and output pathways

Pathways of the Dorsolateral Prefrontal Association Cortex

Four Areas of the Secondary Motor Cortex

Control of movement by motor cortex

•A. microstimulation studies:

- in MI movements of particular contralateral joints (e.g. distal finger) can be elicited by microstimulation;

- in MII contractions of groups of muscles sequentially to produce overall movements of limbs, often bilaterally

Control of movement by motor cortex•B. electrical activity during movement:

- corticospinal neurons active just before initiation of a movement;

- activity related to amount of force necessary to produce the movement;

- directionally-sensitive corticospinal neurons;

- higher-order motor cortex involved in calculating trajectories in space (probably in close communication with cerebellum) and in planning larger-scale movements (probably in close communication with the basal ganglia)

Control of movement by motor cortexC. imaging studies in humans:

- random movements of digits activates MI (precentral gyrus); - planned movements activate MI and supplemental motor cortex;- thinking about planned movements activates supplemental motor

cortex, but not MI.

Of course, this is really too simple…

•Other brain areas involved in movement: - Ventromedial frontal cortex

- involved in body control, posture and whole body movements

- Cerebellum- Basal Ganglia- Brainstem

•In the end, all movement funnels through the alpha motor neuron (final common path)

Motor Hierarchy and Loops

Disorders of the Motor System

•Amyotrophic lateral sclerosis – motor neurons of the brainstem & spinal cord are destroyed. •Huntington’s Disease – progressive destruction of the basal ganglia (GABA).•Muscular Dystrophy – biochemical abnormality affecting the utilization of Ca++ causing wasting away of muscles.•Myasthenia gravis – autoimmune disorder that destroys Ach receptors (starts with head as in drooping eyelids then progresses to swallowing & respiration).•Parkinson’s disease – degeneration of neurons in the striatum due to loss of cells in the substantia nigra that synthesis/release dopamine.

Example: a story of a simple motor system…

Saccadic eye movements

➡ saccades are precise and fast (feedforward)

➡ continual, pervasive adaptation

➡ gaze-direction independent representation needed

DFT, historically, is an account for neurons’ activity

➡“Dynamic neural field” model

Amari, S. Dynamics of pattern formation in lateral-inhibition type neural fields. Biological Cybernetics, 1977, 27, 77-87

Wilson, H. R. & Cowan, J. D. A mathematical theory of the functional dynamics of cortical and thalamic nervous tissue. Kybernetik, 1973, 13, 55-80

Gerstner, Grossberg, Ermentrout, Coombes, Schöner&Spencer, Erlhagen…

“Reaching” task

Distribution of Population Activation (DPA)

precue

responsesignal

500750

time [ms]

movement direction

completeprecue

[Bastian, Riehle, Schöner, 2003]

0 60 120 180 240 300 360

movement direction required in this trial

movement direction

Distribution of population activation =tuning curve * current firing rate3

neurons

[after Bastian, Riehle, Schöner, submitted]

⌧ u(x, t) = �u(x, t) + h+

Zf�u(x0, t)

�!(x� x0)dx0 + I(x, t)

72 Fou n dat ions of Dy na m ic Fi e l d T h eory

activated neurons with overlapping tuning curves, instead of forming only at the receptive field cen-ters of individual neurons. This is shown in Figure 3.7. Figure 3.7b depicts the overlapping recep-tive field outlines for a small sample of neurons, overlaid over the stimulus display. The resulting smooth DPA (computed from all measured neu-rons) during the presentation of a single stimulus can be seen in Figures 3.7c and d.

To obtain the final DPA, an additional nor-malization step is necessary. The neural data stem from a random (and quite limited) sample of neurons from a large population, and one cannot generally assume that the tuning curves of these neurons are distributed equally over the visual space. We may, for instance, have one cluster of neurons in the sample with strongly overlap-ping spatial tunings, such that the correspond-ing region in visual space is overrepresented. Other regions, by contrast, may be covered only sparsely by recorded neurons. An example of this is also visible in the schematic in Figure 3.6, where the space in the central part of the plot is sampled more densely by neurons’ tuning curves. Such uneven sampling can create strong biases in the computed DPA. If we sum the weighted

tuning curves of all neurons, those regions in feature space that are covered by a large number of tuning curves will always tend to produce a high activation value, even if the activity of each individual neuron is relatively low. In contrast, more sparsely sampled regions can never reach very high activation values, even if the individual neurons show strong activity, because very few tuning curves contribute to these activation val-ues. If we assume that the population as a whole represents visual space uniformly, we should compensate for such biases. This is achieved in the study of Jancke and colleagues by divid-ing the weighted sum of tuning curves by the unweighted sum of all tuning curves. This nor-malizes the DPA by scaling the activation up or down according to the density of the sampling at each point in visual space.

We would note that even with this normaliza-tion, the results will not be meaningful if the num-ber of recorded neurons used in the construction of the DPA is too small. In this case, some regions may not be sampled at all by the neurons’ tuning curves. Even though the DPA construction will always yield some activation value for every point in the feature space, these values will not be infor-mative for regions not sufficiently sampled by the recorded neurons. A very small sample size also increases the effects that random noise in the fir-ing rates of individual neurons as well as single neurons with an uncharacteristic response behav-ior have on the resulting activation distribution. Whether the sample of neurons is sufficient cannot be seen directly from an individual DPA—which will always be a smooth distribution of activation over the feature space—but we may judge it by comparing the DPAs produced for different stimu-lus conditions.

Let us now look at the results of the DPA con-struction that Jancke and colleagues obtained for their recordings from cat visual cortex. The elementary stimuli used in the experiment were small squares of light with an edge length of 0.4° of visual angle that were f lashed for 25 ms at one of seven horizontally aligned, equidistant loca-tions at intervals of 0.4° (Figure 3.7a). The DPA analysis was applied to the neural response evoked by these stimuli, using the neurons’ average firing rates over the whole period that a stimulus was presented at each of the seven locations. For all stimuli, the constructed two-dimensional DPAs over visual space show a single, roughly circular

Nasal Temporal

[deg] 2.8º

(d)(c)

(b)Elementary stimuli

FIGURE 3.7: Stimulus conditions and DPA construc-tion in Jancke et al. (1999). (a) Elementary stimuli (0.4º × 0.4º squares of light) were presented at seven hori-zontally shifted positions in the foveal part of the visual field. (b) Receptive field profiles of neurons (gray circles) overlapping and covering the analyzed portion of visual space (black box; gray square illustrates one elementary stimulus). (c) DPA constructed as weighted sum of tuning curves. (d) DPA derived for one elementary stimulus loca-tion overlaid with the stimulus position (small square). Adapted from Jancke et al., 1999.

OUP UNCORRECTED PROOF – FIRSTPROOFS, Mon Aug 03 2015, NEWGEN

01_med_9780199300563_part_1.indd 72 8/3/2015 4:06:00 PM

Perceptual DNF (2D projection) Visual Target DNF

Selected gain map (horizontal)

Selected gain map (vertical)

speed hor.

speed vert.

Saccade generation circuit

Generating precise saccadic eye movements

ICANN 2014, ICDL 2014, Springer 2014

speed hor.

speed vert.

speed hor.

speed vert.

speed hor.

speed vert.

speed hor.

speed vert.

Perceptual DNF

Target DNF

Estimation of error direction

(“too short”, “too long”)

Pan-tilt (propreoception) DNF

Motor memory DNFSaccadic burst generator

where to adapt

how to adapt

Adaptive gains

gain selection

trajectory generation

End-of-Saccadenode (CoS)

panspeed

tiltspeed

gain selection

Fixationnode

End-of-fixationnode

Saccade adaptation

⌧G(x, y, p, t) = con · errsign · T (x, y) ·M(p, t)

Results: Learning the Gain Map

Gain maps for pan Gain maps for tiltSnapshot 1: before first gaze

Snapshot 2: after gaze 1

Learning sensorimotor maps in simulations

0300-30

0300-30Pan gaze angle Pan gaze angle

A: Gain map for horizontal gaze shifts B: Gain map for vertical gaze shifts

Maps are actually 4D

➡Depend on the retinal position of the target and the initial position of the eye

Adaptation of the Gain Maps

What about memory saccades?

RetinaltargetDNF

smoothpursuit

dynamics

Gaze-color

memory

centervis.field

Retinalmultipeak

Serial Order WM

Velocity signal

Gain Maps

Gaze memory

ErrorEstimation Motor command

integration

Gaze Exploration

Gaze Planning(multipeak)

Serial Order WM

Gaze TargetDNF

DifferenceCalculation

Gaze difference

visual perception

camera

colormemory

current gazeDNF

motors

tionF EoF

resetbuild up

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

−0.75

−0.7

−0.65

−0.6

−0.55

−0.5

−0.45

−0.4

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

−0.75

−0.7

−0.65

−0.6

−0.55

−0.5

Double-step saccadesGaze-based target representation

Retinotopic target representation

Presentation of targets Camera movement

Learning to look on a robot

Exploring a scene on a robot

Storck et al, 2014a,b

Learning to reach

• learning the mapping between the gaze-based and body-centred coordinates

Behavioral organization

gaze-direction space

Adaptive mapping

motors: arm

Looking dynamics

camera

motors: head

Arm movement dynamics

proprioceptive space

(1) (2)

(4) (5)

Architecture for Learning to reach

“Move head” “Move hand”

(3) (6)

“Reach and

Learn”

CoS CoS

Behavioral organisation

Lecture motor YS share - Neuroinformaticskiper/Lecture_motor_YS.pdf · 2018-11-19 ·...

Documents

Involuntary Movements

Continual Improvement Process

KLEPTOMANIA INVOLUNTARY STEELING DISORDER

Involuntary Examination

Involuntary TransfersReturns from Leave For Involuntary Transfers and Returns from Leave Wednesday – March 25, 2015

Course Title: Managing Change and Continual Improvement · PT204_Managing Change and Continual Improvement_Information Leaflet V3_May-17 1 Course Title: Managing Change and Continual

Voluntary Involuntary Ecological Illegal

Adams, Involuntary Sins

Smooth Involuntary Cardiac Involuntary Voluntary Skeletal Voluntary Attached to skeleton

CONTINUAL TESTING

Safeguards – Involuntary settlement

Involuntary placement and involuntary treatment of …...Processes of involuntary placement and involuntary treatment of persons with mental health problems can affect the most fundamental

Gangguan Involuntary Movement

Continual Compliance Monitoring

Involuntary Psychiatric Treatment Act

Involuntary Resettlement - Inspection Panel · iv Emerging Lessons Series: Involuntary Resettlement Executive Summary This report on involuntary resettlement is the first in a series

A rating scale that measures INVOLUNTARY … · Abnormal Involuntary Movement Scale ; A rating scale that measures INVOLUNTARY MOVEMENTS known as TARDIVE DYSKINESIA. Tree locations:

Involuntary Conversion

Continual Innovation

Involuntary Resettlement Portfolio Review - World Bankpubdocs.worldbank.org/.../involuntary-resettlement-portfolio-review... · LCR Latin America and Caribbean Region ... Involuntary